New Research Tools Speed Materials Development

Electronics companies are seriously looking at applying the same sort of high-throughput screening processes used to develop new drugs and plastics to get faster results from their own R&D spending.

Start-up Intermolecular (San Jose) has gotten most of the attention from the press lately for its material development tools for the semiconductor sector, which are aimed at enabling fast electrical screening of multiple different isolated sample spots on one substrate instead of having to run research experiments on high-volume manufacturing equipment. But it is not alone in seeing the potential of this approach to speeding up R&D learning cycles. Micron Technology Inc. (Boise, Idaho) recently equipped a lab at the University of Washington (Seattle) with high-throughput combinatorial synthesis and characterization tools, intended for use in the process of its early film development work. Intematix Corp. (Fremont, Calif.) has grown its LED phosphor business using the approach, and is now working on developing materials for energy applications. JSR Corp. (Tokyo) has developed polymers for electronics applications.

Intermolecular’s approach to R&D allows specific chemical solutions to be deposited in 28 isolated, independently programmable cells across a 300 mm wafer.

"Film development is pretty disruptive, even in our R&D fab," explained Scott DeBoer, Micron vice president of process R&D, noting how long it takes just to qualify new materials for fab testing. "We pride ourselves on being able to modify tools pretty rapidly, but it's much easier to do it offline. We can eliminate a new material in the time it would typically take us just to recognize if it is potentially problematic."

Intermolecular's high-throughput tools for semiconductor R&D draw on combinatorial screening technology from Symyx Technologies Inc. (Santa Clara, Calif.), which invested in the start-up to bring the technology to the microelectronics sector. The Tempus line of research systems starts with tools that deposit hundreds of small, isolated spots of different composition across small substrates, and are run through automated characterization and data analysis systems for fast first-pass screening for desired properties. "We're capitalizing on over 10 years experience in other industries, so we didn't have to reinvent the wheel," said Gus Pinto, executive vice president of business development. The more sophisticated fluids-based version for further testing mixes and dispenses chemical solutions in milliliter quantities on 28 isolated, independently programmable cells across a 300 mm substrate . The chemical mixture pulses out through the center of the sample-sized deposition head, and the excess is sucked off at the edge by vacuum. The process is optimized to replicate the characteristics of volume manufacturing equipment, such as single-wafer spin-coaters or electroless deposition tools.

The physical vapor deposition (PVD) systems get similarly controlled composition on isolated sites by using shutters and other masking techniques. One tool enables the user to select among 40 targets in situ in the chamber, without breaking vacuum, to deposit the desired material or alloy on coupons. Tools later in the workflow do progressively more carefully controlled compositions and process steps on full wafers, and can include full wafer processes as well for integration.

Equally important are the characterization systems developed, so far, specifically for each project, based on commercially available metrology tools, to automatically scan across the site-isolated spots on substrates or smaller coupons. Potential cross-contamination in the areas between the isolated sites is reportedly not a problem, because the testing focuses on well-defined isolated sites.

To develop a novel non-volatile memory (NVM) cell for a major chipmaker, Intermolecular modified a customer test chip to enable the company to electrically screen for switching behavior of 5–6 material PVD film stacks directly from confetti-sized samples across a ~3 × 3 in. coupon, without having to send them back into the fab. "That saved four weeks of processing time at the fab," noted David Lazovsky, CEO of Intermolecular, "enabling us to run as many tests in one month as our customer was able to do in a year using conventional processing." He said the project has run 1500 different memory cells through some 12,000 different characterizations so far, including automatic combinatorial testing of algorithms to program the switching of the cell, in about eight months, for what is now the chipmaker's technology of record; it is slated to go into production sometime next year.

The fluids-based workflow has been used to optimize a back-end cleaning chemistry for a semiconductor materials supplier. Intermolecular said it screened different chemistries for cleaning effectiveness and minimal impact on the copper, correlating the physical characterizations with electrical performance in ~10 weeks to develop a new product that is currently in qualification for a 45 nm process at a leading foundry.

This higher-throughput screening technology may also enable the development of radical new materials, such as self-assembled molecular films, that would not be possible otherwise. Intermolecular reported developing a self-assembled monolayer to protect low-k dielectric surfaces in advanced logic from contamination during deposition of a cobalt cap to eliminate the leakage typically resulting from such processes.

Lazovsky noted that the company had to screen some 60 base molecules for the desired properties on copper and low-k dielectrics in ~7600 experiments to find two contenders. Those were not ideal, so the company engineered its own new molecule with the needed properties, then transferred it from a solvent-based to a water-based process to create a masking layer that reportedly significantly reduced leakage on customer's 45 nm wafers without a post clean. The effort took more than 11,500 experiments and 26,000 characterization sets. "For conventional processing, assuming a good team with 24-hour turnaround for each experiment, that would have taken approximately 31 years," Lazovsky said. "Either you'd have to get extremely lucky or you'd give up. But we did it in eight months, with three lots of wafers."

An atomic layer deposition (ALD) process is coming next, according to Lazovsky, and chemical vapor deposition (CVD) is also on the near-term roadmap, although there are no immediate plans for etch. The company is also working on adding electrochemical deposition and chemical mechanical planarization (CMP) processes. It reported that it is now getting ~45% of its revenue from collaborative research projects, 45% from sales of tools, and the remainder from the licensing of IP, although it expects IP to ultimately be a major source.

Micron also sees significant potential in high-throughput initial screenings on dedicated smaller-scale research tools. It recently invested close to $1M through its educational foundation to fund the Micron Laboratory for Combinatorial Materials Exploration at the University of Washington — half in cash and half in equipment. The public lab will focus on applying the combinatorial technology to semiconductor materials, and DeBoer noted that the company influences the materials choices. It supplied the lab with tools for combinatorial PVD, CVD and ALD, a rapid thermal process (RTP) system for the study of maximum thermal gradients, and physical and electrical characterization equipment, including a 2-D X-ray system for collecting multiple types of data in a single scan. "The ALD is still a work in progress," DeBoer said, "but it's important since much of our work is on 3-D applications."

If the physical and lower-level electrical film characteristics look promising, the material is brought back in-house for integration on 300 mm wafers. Although the initial university work is basic research, DeBoer thinks this kind of high-throughput screening will increasingly extend further along into the development process as companies try to cut R&D costs by running as few experiments as possible on 300 mm wafers. "Many people will do this kind of work," he noted, "but the distinguishing factor will be how well they deal with the integration."

Also capitalizing on applying high-throughput combinatorial synthesis development techniques to electronics is Intematix, which has used the deposition and characterization tools it developed to create new kinds of phosphors — particularly yellows that convert blue LED light to white — that avoid the prior art of the existing cerium-doped YAG yellow alternatives. Chris Bajorek, vice president of advanced development, said that the company has signed ~20 customers, and is working with Samsung (Kyunggi-Do, Korea) on LEDs for display backlights to become a significant player in the market after only two years of targeting the niche. The company identified the phosphor compositions by depositing continuous gradients of films of various compositions in overlapping patterns across a small substrate, but Bajorek said Intematix could now do both gradient and discrete deposition with fluid and gas technologies. For fast characterization, the company developed an automated X-ray microprobe for screening crystal structure and composition, and measuring conductivity and dielectric properties with a non-contact proximity evanescent microwave scanning technology developed by Lawrence Berkeley National Laboratory (Berkeley, Calif.).

The company has now shifted its focus to emerging energy applications. "Some 50% of power is used for lighting, so increasing the efficiency of lighting by 50% by going to LEDs would have far more impact on consumption than solar power," Bajorek noted. He said the company is working with customers on developing materials for solar films (both CIGS and new alternatives) for rechargeable batteries and fuel cells, as well as working with some customers in the semiconductor space. It reported early development work on a lower-cost catalyst system for methanol fuel cells in a design of experiments (DOE)-funded research project using alloys of 10–30% platinum to replace pure platinum on a nanostructure for efficient exposure.